Data-driven enhanced control method for flexible double-arm humanoid robot

Abstract

A data-driven enhanced control method for a flexible double-arm humanoid robot belongs to the technical field of robot motion control, and comprises the following steps: step S1, based on motion information of a flexible robot, converting a system model containing unknown kinetic parameters into a linear discrete time data model with equivalent characteristics; s2, designing an improved projection estimation algorithm to estimate pseudo partial derivative parameters of the robot system in real time; s3, designing a finite time performance exponential function, and obtaining a data driving enhancement controller based on an optimal control principle; and S4, constructing a data-driven high-order control barrier function based on the linear time data model, and obtaining a data-driven control law which simultaneously meets stability and preset control performance by solving a quadratic programming problem. The modeling steps are effectively simplified, the modeling efficiency is improved, and the finite time convergence performance and the optimal control performance of the flexible double-arm robot are both considered.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of robot motion control, and specifically provides a data-driven enhanced control method for a flexible dual-arm humanoid robot. BACKGROUND

[0002] With the continuous progress of automation technology, the application field of intelligent robots is increasingly wide, and its control task is becoming more and more complex. Flexible humanoid robots can simulate human behavior and action and complete various complex tasks, so they show great value and potential in research and application. However, the dual-arm motion control problem of humanoid robots has always been a major challenge in the field of robot motion control.

[0003] The dynamics system of a flexible dual-arm robot is particularly complex due to its highly nonlinear and strongly coupled characteristics, which poses significant challenges to dynamics modeling and control. Traditional control strategies usually need to rely on accurate parameter setting and extensive experimental debugging, which not only consumes time but also costs high. In addition, the robot flexible joint is prone to excite structural resonance when accelerating or decelerating or the external load changes, which in turn induces mechanical vibration. This not only affects the accuracy of trajectory tracking, but also introduces additional disturbance to the control system, significantly increasing the difficulty of maintaining high-precision motion control while ensuring the overall stability of the system. Therefore, it is particularly important to explore data-driven control strategies in the case of incomplete or completely unknown dynamics information of a flexible dual-arm humanoid robot.

[0004] It should be noted that traditional model predictive control algorithms highly depend on accurate robot mathematical models, although they can model the dynamic characteristics of the system under certain assumptions, but this is very time-consuming and relatively complex, making it difficult to directly analyze the controller. In addition, existing robot control methods are mostly based on asymptotic convergence framework, whose convergence speed is difficult to guarantee. This to some extent restricts the motion performance of the flexible dual-arm robot, especially when the system model has uncertainty, the problem is more prominent. SUMMARY

[0005] In order to overcome the shortcomings of the prior art, the present application provides a data-driven enhanced control method for a flexible dual-arm humanoid robot, which effectively simplifies the modeling step and improves the modeling efficiency, and takes into account the finite time convergence performance and optimal control performance of the flexible dual-arm humanoid robot. First, a dynamic model of the flexible robot is established, and a linear discrete-time data model with equivalent properties is established in combination with real-time feedback of motion control information. Second, in each control cycle, an improved projection estimation algorithm is used to estimate the pseudo-derivative parameters of the robot system online in real time. On this basis, adjustable step parameters and penalty factors are introduced to enhance the adaptability and flexibility of the estimation algorithm under different control tasks. Finally, based on the constructed linear discrete-time data model, a finite time performance index function and a data-driven high-order control barrier function are designed, and a data-driven enhanced control law that guarantees system stability and meets the preset control performance is generated by solving a quadratic programming optimization problem.

[0006] In order to achieve the above-mentioned purpose, the technical scheme of the present application is as follows: A data-driven enhanced control method for a flexible dual-arm humanoid robot, comprising the following steps: Step S1, based on the motion information of the flexible robot, the system model containing unknown dynamic parameters is converted into a linear discrete-time data model with equivalent properties; Step S2, an improved projection estimation algorithm is designed to estimate the pseudo-derivative parameters of the robot system in real time; Step S3, a finite time performance index function is designed, and a data-driven enhanced controller is obtained based on the optimal control principle; Step S4, a data-driven high-order control barrier function is constructed based on the linear time data model, and a data-driven control law that satisfies stability and preset control performance is obtained by solving a quadratic programming problem.

[0007] Further, in step S1, first, a dynamic model of the flexible dual-arm humanoid robot system in joint space is established; second, a parameter projection estimation algorithm is designed using system motion control information to estimate the pseudo-derivative parameters of the system online , and an equivalent linear time data model is established.

[0008] In step S1, the following dynamic model of the flexible joint robot is constructed: (1) ; Wherein, represents the unknown inertia matrix, the Coriolis matrix and the gravity matrix of the robot, represents time, is the control input of the system at time t, represents the robot system Position, velocity, and acceleration at any given moment; Based on the real-time motion control information of the system, a discrete linear time data model is established: (2); in, Represents robots Position output at any given moment Represents robots Time-based control input, These are pseudo-partial derivative parameters for the robot's discrete-time system, used to construct the robot's dynamic data model.

[0009] Furthermore, in step S2, in each control cycle, based on the robot's control input... and control output We design an estimation criterion function for the pseudo-partial derivative parameter and obtain the estimation formula by solving for the extrema. Based on this, we design an adjustable step size parameter and a penalty factor to enhance the flexibility of the estimation algorithm.

[0010] In step S2, the following estimation criterion function is constructed: (3); in, Represents robots Position output at any given moment Yes The estimated value, Yes The estimated value, It is a penalty factor that makes the estimated value smoother. , Represents robots Time-based control input; After minimization, the following estimation formula is obtained: (4); in, This is the step size parameter used to adjust the estimation algorithm.

[0011] Furthermore, in step S3, based on the linear time data model, a finite-time performance exponential function is constructed by designing convergence ratio parameters; on this basis, the data-driven control law is solved online using the optimal control principle, so as to enhance the rapid convergence of the robot system without relying on detailed system dynamics information.

[0012] In step S3, the following finite-time performance function is designed: (5); in, is a designed robot convergence ratio parameter for adjusting the convergence speed of the controller, is a designed weight parameter for limiting the variation of the control input signal, so that the control input is more smooth, is the tracking signal at time t; According to the optimal control principle, the following data-driven enhanced control law is obtained: (6); wherein, is a designed control parameter, so that the robot controller is more general.

[0013] In the step S4, according to the performance constraint function, a high-order data-driven control barrier function with order is constructed to ensure the error performance constraint with higher degree of freedom; on this basis, by solving the quadratic programming problem based on the data-driven enhanced controller and the high-order control barrier function, the data-driven control law that meets the stability and the preset control performance at the same time is generated in real time.

[0014] In the step S4, for the flexible joint robot, the following data-driven control barrier function is constructed: (7); wherein, is a symbol abbreviation of , , , is a K-type function, , is an error boundary, is an error performance constraint function, is the tracking signal at time t; Finally, by solving the quadratic programming problem of fusing the control function and the data-driven model, the optimal control input is generated in real time: (8); wherein, are respectively the upper and lower limits of the control input, is the nominal control input obtained by formula (6), is the optimal control input.

[0015] The technical concept of the present application is: first, a dynamic model of the flexible robot is established, and a linear discrete-time data model with equivalent properties is established combined with real-time feedback of motion control information; second, in each control period, an improved projection estimation algorithm is used to perform online real-time estimation of the pseudo partial derivative parameters of the system; on this basis, adjustable step parameters and penalty factors are introduced to enhance the adaptability and flexibility of the estimation algorithm under different control tasks; finally, based on the constructed linear discrete-time data model, a finite time performance index function and a data-driven high-order control barrier function are designed, and by solving a quadratic programming optimization problem, a data-driven enhanced control law is generated which guarantees the stability of the system and meets the preset control performance.

[0016] Compared with the prior art, the beneficial effects of the present application mainly manifest in: 1) The present application establishes a linear discrete-time data model for high-precision motion control of a flexible dual-arm robot. Compared with the complex and time-consuming modeling process in traditional control methods, the method of the present application effectively simplifies the modeling steps and improves the modeling efficiency, providing a more efficient and accurate control framework for high-precision motion of the flexible dual-arm robot.

[0017] 2) Compared with traditional asymptotic convergence control methods, the present application takes into account both the finite time convergence performance and the optimal control performance of the flexible dual-arm robot. By designing a finite time performance index function, the system is ensured to be quickly stabilized within a specified time, and a data-driven high-order control barrier function is established to strictly ensure that the system meets the error constraint requirements throughout the running process. BRIEF DESCRIPTION OF DRAWINGS

[0018] Fig. 1 is the tracking control effect comparison curve of the left arm joint of the robot in the present application; Fig. 2 is the tracking control effect comparison curve of the right arm joint of the robot in the present application; Fig. 3 is the data-driven enhanced control method flowchart for the flexible dual-arm humanoid robot in the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical scheme and advantages of the present application clearer and more apparent, the present application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0020] Referring to Figs. 1-3 A data-driven enhanced control method for a flexible dual-arm humanoid robot, comprising the following steps: Step S1, based on the motion information of the flexible robot, converts the system model containing unknown kinetic parameters into a linear discrete-time data model with equivalent characteristics; In step S1, first, a dynamic model of the flexible dual-arm humanoid robot system in joint space is established; second, a parameter projection estimation algorithm is designed using system motion control information to estimate the pseudo partial derivative parameters of the system online An equivalent linear time data model is established, and the process is as follows: The following dynamic model of the flexible joint robot is constructed: (1) ; Wherein, represent the unknown inertia matrix, the Coriolis force matrix and the gravity matrix of the robot, denotes time, is the control input of the system at time t, denotes the position, velocity and acceleration of the robot system at time t; 1.2) Further, a discrete linear time data model is established according to the real-time motion control information of the system: (2) ; Wherein, denotes the position output of the robot at time t, denotes the control input of the robot at time t, is the pseudo partial derivative parameter of the robot discrete-time system, is the pseudo partial derivative parameter of the robot discrete-time system, and its initial value is set to .

[0021] Step S2, an improved projection estimation algorithm is designed to estimate the pseudo partial derivative parameters of the robot system in real time; In step S2, in each control period, based on the control input and the control output of the robot, an estimation criterion function about the pseudo partial derivative parameters is designed, and the estimation formula is obtained by solving the extreme value operation, on this basis, adjustable step size parameters and penalty factors are designed to enhance the flexibility of the estimation algorithm; the process is as follows: The following estimation criterion function is constructed: (3) ; Wherein, , denotes the control input of the robot at time t, denotes the control input of the robot The position output at the time instant, is an estimate of , is an estimate of , a penalty factor is set as ; (4) ; wherein a step parameter of the estimation algorithm is set as ; A reset algorithm is designed to enhance the stability of the robot system: (9) wherein a threshold parameter is set as .

[0022] Step S3, a finite-time performance index function is designed, and a data-driven enhanced controller is obtained based on the optimal control principle; In the step S3, based on the linear time data model, a finite-time performance index function is constructed by designing a convergence ratio parameter; on this basis, based on the optimal control principle, the rapid convergence of the robot system is enhanced without relying on the detailed information of the system dynamics; the process is: 3.1) design the following finite-time performance function: (5) ; wherein a convergence ratio parameter is set as , and a weight parameter is set as ; 3.2) according to the optimal control principle, the following data-driven enhanced control law is obtained: (6) ; wherein a controller parameter is set as .

[0023] Step S4, a data-driven high-order control barrier function is constructed based on the linear time data model, and a data-driven control law that meets the stability and preset control performance at the same time is obtained by solving a quadratic programming problem: In the step S4, according to the performance constraint function, a high-order data-driven control barrier function with order is constructed to ensure the error performance constraint with higher freedom; on this basis, by solving a quadratic programming problem based on the data-driven enhanced controller and the high-order control barrier function, a data-driven control law that meets the stability and preset control performance at the same time is generated in real time, and the process is: 4.1) For a flexible joint robot, construct the following data-driven control barrier function: (7); wherein, is a symbolic abbreviation of , , , is a K-type function, , ; is an error boundary, ; is an error performance constraint function, is a tracking signal at the moment ; Finally, by solving a quadratic programming problem of fusing the control function and the data-driven model, the optimal control input is generated in real time: (8); wherein, are the upper and lower limits of the control input, is the nominal control input obtained from equation (6), is the optimal control input.

[0024] In the experiment of the embodiment, the sampling period T of the robot is set to , that is, the equivalent dynamic data model is constructed using every input and output information obtained from the system, and then the control input is solved and sent to the robot. The comparison curve of the tracking control effect of the left arm joint of the robot is shown in Fig. 1 . The comparison curve of the tracking control effect of the right arm joint of the robot is shown in Fig. 2 . From the experimental results, it can be seen that compared with the traditional model predictive control algorithm, the proposed data-driven enhanced controller has better steady-state accuracy and transient response. In the entire control process, the motion trajectory of the robot under the application is relatively smooth, and there is no overshoot, oscillation and other phenomena.

[0025] The above describes only the preferred specific embodiments of the application, but the protection scope of the application is not limited thereto, and any skilled person in the art can make equivalent replacements or changes to the technical solutions and inventive concepts of the application within the technical range disclosed by the application, which should be covered within the protection scope of the application.

Claims

1. A data-driven augmented control method for a flexible dual-arm humanoid robot, characterized in that, The method includes the following steps: Step S1: Based on the motion information of the flexible robot, the system model containing unknown dynamic parameters is transformed into a linear discrete-time data model with equivalent properties; Step S2: Design an improved projection estimation algorithm to estimate the pseudo-partial derivative parameters of the robot system in real time; Step S3: Design a finite-time performance exponential function and obtain a data-driven enhanced controller based on the optimal control principle; Step S4: Construct a data-driven high-order control barrier function based on a linear time data model, and obtain a data-driven control law that simultaneously satisfies stability and preset control performance by solving a quadratic programming problem.

2. The data-driven enhanced control method for a flexible dual-arm humanoid robot as described in claim 1, characterized in that, In step S1, firstly, a dynamic model of the flexible dual-arm humanoid robot system in the joint space is established; secondly, a parameter projection estimation algorithm is designed using the system motion control information to estimate the pseudo-partial derivative parameters of the system online. Establish an equivalent linear time data model.

3. The data-driven enhanced control method for a flexible dual-arm humanoid robot as described in claim 2, characterized in that, In step S1, the following dynamic model of the flexible joint robot is constructed: （1）； in, These represent the robot's unknown inertia matrix, Coriolis force matrix, and gravity matrix, respectively. Indicates time, For the system Time-based control input, Representing robot system Position, velocity, and acceleration at any given moment; Based on the real-time motion control information of the system, a discrete linear time data model is established: （2）； in, Represents robots Position output at any given moment Represents robots Time-based control input, These are pseudo-partial derivative parameters for the robot's discrete-time system, used to construct the robot's dynamic data model.

4. A data-driven enhanced control method for a flexible dual-arm humanoid robot as described in any one of claims 1 to 3, characterized in that, In step S2, during each control cycle, based on the robot's control input... and control output We design an estimation criterion function for the pseudo-partial derivative parameter and obtain the estimation formula by solving the extremum operation. Based on this, we design an adjustable step size parameter and a penalty factor to enhance the flexibility of the estimation algorithm.

5. The data-driven enhanced control method for a flexible dual-arm humanoid robot as described in claim 4, characterized in that, In step S2, the following estimation criterion function is constructed: （3）； in, Represents robots Position output at any given moment Yes The estimated value, Yes The estimated value, It is a penalty factor that makes the estimated value smoother. , Represents robots Time-based control input; After minimization, the following estimation formula is obtained: （4）； in, This is the step size parameter used to adjust the estimation algorithm.

6. A data-driven enhanced control method for a flexible dual-arm humanoid robot as described in any one of claims 1 to 3, characterized in that, In step S3, based on the linear time data model, a finite-time performance exponential function is constructed by designing convergence ratio parameters. On this basis, the data-driven control law is solved online using the optimal control principle to enhance the rapid convergence of the robot system without relying on detailed system dynamics information.

7. The data-driven enhanced control method for a flexible dual-arm humanoid robot as described in claim 6, characterized in that, In step S3, the following finite-time performance function is designed: （5）； in, These are the convergence ratio parameters of the designed robot, used to adjust the convergence speed of the controller. These are weighting parameters designed to limit variations in the control input signal, resulting in a smoother control input. yes Real-time tracking signal; Based on the optimal control principle, the following data-driven enhanced control law is obtained: （6）； in, These are the control parameters designed to make the robot controller more general.

8. A data-driven enhanced control method for a flexible dual-arm humanoid robot as described in any one of claims 1 to 3, characterized in that, In step S4, a performance constraint function is constructed based on the performance constraint function. Higher-order data-driven control barrier function To ensure error performance constraints with higher degrees of freedom, a data-driven control law that simultaneously satisfies stability and preset control performance is generated in real time by solving a quadratic programming problem that unifies the data-driven enhanced controller and the high-order control barrier function.

9. The data-driven enhanced control method for a flexible dual-arm humanoid robot as described in claim 8, characterized in that, In step S4, for the flexible joint robot, the following structure is constructed: Level data-driven control barrier function: （7）； in, Yes abbreviation of symbols, , , It is a K-type function. , It is the error boundary. It is the error performance constraint function. yes Real-time tracking signal; Finally, by solving a quadratic programming problem that integrates the control function and the data-driven model, the optimal control input is generated in real time. （8）； in, These are the upper and lower limits of the control input, respectively. The nominal control input is obtained from equation (6). This is the optimal control input.

Images